Accelerating the Design of Sustainable Concrete with CDW Through Adaptive Experiments and Machine Learning for Conventional and Additive Manufacturing Applications

Abstract

This study presents an integrated computational framework combining machine learning (ML), Bayesian optimization (BO), and experimental validation to design sustainable concrete incorporating construction and demolition waste (CDW) for both conventional and 3D concrete printing (3DCP) applications. In traditional mortar, ML models accurately predict 28-day compressive strength using early-age results, enabling nearly 10× acceleration of mixture design cycles. Optimized mixtures achieved up to 50 % reduction in ordinary Portland cement (OPC) while maintaining strengths above 40 MPa. For 3DCP, a multi-objective BO framework simultaneously maximized buildability and CDW content. Experimental results validated enhanced buildability (up to 10 printed layers) and compressive strengths up to 60 MPa in cast samples and 52 MPa in 3D printed samples, even with up to 97 % CDW. The study highlights the potential of data-driven methods to transform sustainable material design in the construction industry.